In the design of distributed and complex systems, as well as in the run-time process, the service-based model plays a central role. Under the assumption of teleological reasoning, each service can be related to a goal. Therefore, services are described by a functional model which relies on a core reasoning approach. The design process involves the composition of subservices, the fusion of services and to check if the composition and/or fusion satisfies the requirements. Today this process is still largely conducted by hand, and leads to a very time-consuming process which often generates mistakes. As a consequence, it has become a bottleneck in many information sharing applications. In order to automate this process, we propose in this paper a methodology which deals with these insufficiencies on two steps. The first is the design one, where our approach is based on the Information Flow (IF) model which makes possible a mapping process of goal ontologies. In the run-time step, the specification involves a Multi Agent System (MAS). We propose an algorithm describing the mechanism of the dynamic mapping of services, in which the intelligent agents are able to map dynamically the distant goal ontologies.